The Availability of Digital Evidence

The power of social media to enable people to organize and share information during conflict can incentivize repressive governments to restrict—or shut down—access to the Internet to control what people are saying and sharing online. This can be implemented both by a country's own government or by an occupying power. For example, during the violent conflict in the Tigray region of Ethiopia between November 2020 and 2022,Footnote ¹ the region was subject to an Internet shutdown impacting more than six million people and restricting information emerging from the conflict.Footnote ² This two-year information black hole prevented impacted communities from organizing and sharing stories of the harms committed against them, and will impact how this conflict is memorialized and the kinds of accountability projects that are pursued in the future.

Following Russia's unlawful invasion of Ukraine in early 2022, the city of Mariupol was subjected to a two-month siege and information blackout,Footnote ³ deliberately caused by Russian forces shelling the last cell tower.Footnote ⁴ Approximately 100,000 people remained in the city, with hundreds of thousands of others having already fled. As the only international journalists left in the city, reporters for the Associated Press continued to document and report by smuggling out memory cards of data from the region, at great personal risk,Footnote ⁵ without which we would not know the extent of the atrocities by Russian forces. Since the October 7, 2023 unlawful attack by Hamas on Israel, Israel has instituted a “complete siege” of Gaza.Footnote ⁶ This has included Internet shutdowns and information blackouts for 2.2 million people,Footnote ⁷ driven by the destruction of communications infrastructureFootnote ⁸ and shortages of fuel and power,Footnote ⁹ which has impeded the ability of the people of Gaza to communicate with one another and to document and share what is happening to them.Footnote ¹⁰

Whether digital information can be used to monitor and seek accountability for laws of war violations depends on the availability of information. 2022 saw 187 Internet shutdowns in thirty-five countries, marked by “the weaponization of shutdowns during armed conflict.”Footnote ¹¹ Information blackouts during conflict can “provide cover for atrocities and breed impunity,”Footnote ¹² as the world is unable to see the extent of the atrocities being committed on the ground in real time. Power blackouts mean that devices go uncharged and valuable digital evidence for future accountability processes is not captured.

Another way in which access to information uploaded by users in conflict zones is impacted is through social media platforms’ content moderation policies, which can result in the removal, flagging, or deprioritization of content.Footnote ¹³ For example, two weeks into the Russia-Ukraine war, YouTube promoted that it had removed more than 15,000 videos for violating their terms of service.Footnote ¹⁴ As Human Rights Watch has reported, this “prevents the use of that content to investigate people suspected of involvement in serious crimes, including war crimes.”Footnote ¹⁵ While there may be a legitimate public interest in removing content that could be harmful to users, it is imperative that platforms make the content available to researchers and investigators that need it to identify instances of international law violations during conflict. Additionally, social media platforms can use their algorithms to deprioritize content and to silence the voices of some parties to conflict. At the time of writing, the voices of Palestinians and supporters are being silenced through “content removals, suspension or deletion of accounts, inability to engage with content, inability to follow or tag accounts, restrictions on the use of features such as Instagram/Facebook Live, and ‘shadow banning’ [making a user's content no longer visible to others].”Footnote ¹⁶ Through these measures, platforms are making it harder for people to tell their stories and creating challenges for war crimes investigators to identify potential violations.

When digital evidence is available, it can be difficult to find useful data through the noise of social media. For example, in 2022, 6,000 tweets were posted to Twitter every secondFootnote ¹⁷ and more than 500 hours of video are currently uploaded to YouTube every minute. With that level of usage, discovering relevant content that shows potential violations of international law, or helps to identify alleged victims or perpetrators, is incredibly difficult. Machine learning has huge potential to help with the process of identifying and classifying relevant content and, as the technology improves, will greatly assist war crimes investigators in their monumental challenge of evidence collection and analysis. But machine learning must be implemented cautiously, so as to avoid the perpetuation of bias and discrimination inherent in many technology products.Footnote ¹⁸

The Reliability of Digital Evidence

Even where digital information is available, rampant disinformation and deepfakes can call the reliability of content into question. It is well documented that the Internet is replete with disinformation, particularly in the context of electionsFootnote ¹⁹ and COVID-19.Footnote ²⁰ Disinformation during warfare is nothing new, but the speed with which disinformation can spread using social media is unprecedented and may affect the reliability of digital evidence, or at least the perception of its reliability. Two days after Hamas's attack on Israel, David Gilbert at Wired reported that the Israel-Hamas war “is drowning X in disinformation.”Footnote ²¹ Some of that was driven by recent changes made to the X (formerly Twitter) platform. Some was driven by users posting content from other contexts: in some instances video was recycled from other conflicts, including Syria in 2020,Footnote ²² in others scenes from a video game were represented as Hamas attacks on Israel.Footnote ²³ Additional unverified claims have circulated online, including much speculation about who is responsible for the hospital blast at Al-Ahli Baptist Hospital.Footnote ²⁴ In the Russia-Ukraine conflict, Russia has deliberately engaged “systemic information manipulation and disinformation by the Kremlin . . . as an operational tool in its assault on Ukraine”Footnote ²⁵ to advance its revisionist history propaganda campaign and to tout the success of its military efforts to garner support at home.Footnote ²⁶

In addition, we have witnessed the explosion of generative AI over the past year, which “can produce various types of content, including text, imagery, audio and synthetic data.”Footnote ²⁷ The most mainstream is currently text-based Chat GPT, but there are steadily more user-friendly AI that can create increasingly realistic photos and videos, such as Dall-E, Craiyon, Deep AI, and Runway. Deepfake technology poses significant challenges to the reliability of user generated content, as it contributes to the “truth decay” of our information ecosystem.Footnote ²⁸

Disinformation and deepfakes help to fuel false narratives and whip up support for parties who manipulate the truth for their own gains. Most of us are not yet equipped with the visual verification techniques necessary to identify fake posts, which can be particularly dangerous in a conflict situation, where inflammatory content may lead to increased violence on the ground. Several platforms, such as Facebook, now have independent fact-checkers to help review content uploaded by users,Footnote ²⁹ and many media outlets, such as the New York Times, will provide fact checking services, but they do not catch everything and enthusiasm for these efforts may be waning.Footnote ³⁰ Against this backdrop, digital investigators must carefully authenticate content. This is the process by which content is geolocated and chronolocated to ensure that a post is what it claims to be, and not something that is being taken out of context or repurposed. Ensuring content is methodically analyzed in this way will help to dispel some of the concerns with using it in war crimes prosecutions.

The Admissibility of Digital Evidence

Once investigators have discovered relevant digital content of potential war crimes and determined its reliability, the question of its admissibility will need to be addressed in each jurisdiction in which its submission is sought. A primary concern is establishing the chain of custody of digital evidence, which is vital to many criminal justice systems and requires demonstration of where and when evidence was received and stored. Luckily, there has been significant evolution in this space over the past decade: there are now advanced web capture tools like Hunch.ly and digital archives, such as those created and maintained by Mnemonic, to forensically store and secure content. In addition, the Berkeley Protocol on Digital Open-Source Investigations outlines methods to effectively preserve chain of custody and to create minimum standards of practice for digital investigators seeking to gather evidence of international crimes.Footnote ³¹ Recently, the International Criminal Court (ICC) has launched its own evidence submission platform, OTPLink, to provide users with a “seamless and secure method for submitting potential evidence.”Footnote ³² All of these innovations allow us to place increasing trust in the chain of custody of digital evidence documenting international law violations.

Using the ICC as an example, generally, the court takes a liberal approach to the admission of evidence.Footnote ³³ It has relied heavily on publicly available information to build its cases since the court's inception,Footnote ³⁴ and since the early 2010s has increasingly utilized digital evidence, particularly user-generated content, such as social media posts. Beginning in 2013, the ICC relied on Facebook to establish a relationship between parties involved in alleged witness tampering in the case against Jean-Pierre Bemba Gombo.Footnote ³⁵ In the 2016 case of Ahmad Al-Faqi Al-Mahdi, a range of digital information was used to help to secure Al-Mahdi's guilty plea and his conviction as a co-perpetrator for the war crime of destruction of cultural property.Footnote ³⁶ In 2017, the Court relied upon videos uploaded to Facebook that showed the Libyan General Mahmoud Mustafa Busayf al-Werfalli directly involved in the killing of thirty-three persons in order to issue an arrest warrant for him.Footnote ³⁷

The biggest test for the use of digital evidence at the ICC, and particularly user-generated content, is likely to come from cases stemming from Russia's war with Ukraine. Since the armed conflict broke out in February 2022, millions of pieces of content have been collected documenting alleged violations. Groups such as Bellingcat are tracking a wide variety of harms,Footnote ³⁸ and Mnemonic has built a digital archive to forensically store evidence.Footnote ³⁹ How exactly the Office of the Prosecutor will rely on this information in its investigation and any resulting cases remains to be seen, but it is likely to play a significant role as ICC Chief Prosecutor Karim Khan has emphasized the importance of collecting digital materials in this and other situations.Footnote ⁴⁰

Digital evidence can be highly probative as evidence of the necessary intention and commission of international crimes. However, as this essay has shown, numerous challenges can arise in its collection. Even when digital evidence is available, it is not a silver bullet; we are unlikely to have a video showing the entire chain of an event, including the perpetrator's role in its commission. But we can use user generated content to corroborate other evidence, to provide contextual information about the conflict, to lead us to potential witnesses who may be able to provide testimony, or as linkage evidence to establish connections and chains of command between alleged perpetrators.Footnote ⁴¹

Concluding Observations

To achieve the potential of digital evidence to transform the accountability landscape, we must continue to grapple with how to enhance its availability, reliability, and admissibility before courts. Social media platforms must be required to make potential evidence of war crimes available to investigators and must commit to not silencing voices online. Platforms must also do a better job of identifying disinformation and AI generated deepfakes and placing warnings on posts that are spreading falsehoods. We must harness the power of AI for good and use machine learning to help investigators to identify and analyze relevant content. The ICC is already deploying AI to help with evidence pattern analysis through its digitization project, Project Harmony.Footnote ⁴² This and other initiatives should be supported by governments, big tech, and civil society to create a digital infrastructure capable of harnessing the power of user-generated content to pursue accountability for international crimes.